IBM SPSS Collaboration and Deployment Services 8.3 RESTful Scoring Service Developer's Guide
IBM SPSS Collaboration and Deployment Services 8.3 includes a RESTful Scoring Service that lets client applications use real-time scores from predictive models developed in IBM SPSS Modeler, IBM SPSS Statistics, or third-party tools. The service fetches a specified model, loads it, invokes the correct scoring implementation, and returns the result. Scoring can be synchronous or asynchronous, and the service can load multiple models at the same time.
This developer's guide explains how clients access the service over HTTP or HTTPS, specify request URLs, headers, and JSON request bodies, and receive response headers and bodies. It provides an API reference for the service, configuration, metadata, score, metric item, and metric value resources, plus a JSON reference for the related objects and elements.
| Product version | Version 8, release 3, modification 0 |
|---|---|
| Supported model types | Regression (linear and logistic), decision trees, decision lists, neural networks, and naïve Bayes |
| Scoring modes | Synchronous or asynchronous |
| Base URL | http://{server}:{port}/scoring/rest |
| Authentication | HTTP basic access authentication using a Base64-encoded user name and password |
The base URL is http://{server}:{port}/scoring/rest, where {server} is the server name or IP address and {port} is the repository port.
It uses HTTP basic access authentication, with the Authorization header containing a Base64-encoded user name and password.
It supports regression (linear and logistic), decision trees, decision lists, neural networks, and naïve Bayes defined in IBM SPSS Modeler streams or PMML from IBM SPSS Statistics.
Yes. Scoring can be synchronous or asynchronous, depending on whether the client needs to wait for a score.